Hi
@araj! First of all you need to understand that Data Abstraction is a tool of Computer Science and not of Scheme. The book has used Scheme, so that will the reference for answering your question.
Just like the book (SICP) states, before coming to Data Abstraction the book talks about Procedural Abstraction: We assume that we have smaller procedures which do smaller tasks for us. With that assumption we define our main procedure and look into the details of implementing those smaller programs later. Same is the case for Data Abstraction. We need not use it for simple data. Not that we can't, but we don't. However, when dealing with complex forms of data, their manipulation could be tricky. The book explains this with an example of rational numbers. Now, we surely know that addition of rationals is trickier than that of integers. So, if you look at the section on Data Abstraction in the book, you will note that without worrying about the details, the strategy assumes that we already have procedures which
0. create a rational
1. retrieve the numerator/denominator of the rational
Making use of these assumptions, they have simply defined procedures for adding/subtracting etc of rationals.
And then looked into defining the procedures make-rat, numer and denom later on.
NOTE THAT they didn't worry about how will they make/store a rational number or access its numerator/denominator before defining the add-rat, sub-rat, mul-rat etc...
The BIG thing really in all this is 'Abstraction' itself: whether procedural or data. The underlying idea is to assume that you have someone to do the small tasks for you, and using all those small result to get your net result.
Hope that answers your questions!
Any further questions are welcomed.